In order to quickly and accurately monitor crop growth, winter wheat was used as research object, and UAV hyperspectral images of different growth stages were acquired. Firstly, the hyperspectral data of UAV were used to construct the spectral index, and the indices of four growth stages were analyzed respectively, which were related to the biomass, leaf area index and the new growth monitoring indicator (GMI) constructed by the two physiological parameters of biomass and leaf area, and then a single exponential regression model was established with four spectral indices that were strongly correlated with GMI, and GMI inversion models of winter wheat growth stages were established by using three machine learning methods: multiple linear regression, partial least square and random forest. Finally, the best model was applied to the UAV hyperspectral image to obtain the growth monitoring map. The results showed that the correlation between the spectral index and GMI of winter wheat was high, and most of the indices reached significant levels. The correlation between NDVI, SR, MSR and NDVI×SR and GMI was higher than that of biomass, leaf area index and GMI. The regression model established by the single spectral index of each growth stage, the best performing model corresponding to the spectral indices were NDVI×SR, NDVI, SR, NDVI and NDVI×SR; compared with GMI inversion model constructed by three methods, the flowering stage model MLR-GMI had the best effect. The model modeling R2, RMSE and NRMSE of this stage were 07164, 00963 and 1590%, respectively.